Fix rearrange-pages DUPLICATE producing shared page nodes (pypdf cyclic-references CI break) (#6851)

This commit is contained in:
Anthony Stirling
2026-07-01 13:40:27 +01:00
committed by GitHub
parent c22ecc6c09
commit 9d3701a585
2 changed files with 45 additions and 2 deletions
@@ -3,9 +3,12 @@ package stirling.software.SPDF.controller.api;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Collections;
import java.util.HashSet;
import java.util.List;
import java.util.Locale;
import java.util.Set;
import org.apache.pdfbox.cos.COSDictionary;
import org.apache.pdfbox.pdmodel.PDDocument;
import org.apache.pdfbox.pdmodel.PDPage;
import org.apache.pdfbox.pdmodel.PDPageTree;
@@ -261,10 +264,19 @@ public class RearrangePagesPDFController {
log.info("newPageOrder = {}", newPageOrder);
log.info("totalPages = {}", totalPages);
// Snapshot the desired pages before mutating the source document's page tree.
// Snapshot desired pages before mutating the tree; clone repeats (e.g. DUPLICATE)
// so each slot is a distinct node, not one PDPage under multiple /Kids.
List<PDPage> newPages = new ArrayList<>(newPageOrder.size());
Set<Integer> seenIndices = new HashSet<>();
for (Integer idx : newPageOrder) {
newPages.add(document.getPage(idx));
PDPage page = document.getPage(idx);
if (!seenIndices.add(idx)) {
// Duplicate index: distinct page node sharing content/resources.
COSDictionary clonedDict = new COSDictionary();
clonedDict.addAll(page.getCOSObject());
page = new PDPage(clonedDict);
}
newPages.add(page);
}
// Rearrange in-place on the source document rather than copying pages into a
@@ -9,6 +9,7 @@ import java.io.File;
import java.io.IOException;
import java.nio.file.Files;
import java.util.ArrayList;
import java.util.HashSet;
import java.util.List;
import org.apache.pdfbox.Loader;
@@ -302,6 +303,11 @@ class RearrangePagesPDFControllerTest {
assertNotNull(response);
// 2 pages * 3 duplicates = 6 final pages
assertEquals(6, realDoc.getNumberOfPages());
// Each duplicate must be a distinct page node in the saved output; a shared
// node under multiple /Kids is an invalid tree readers reject as cyclic.
List<Object> savedPages = reloadAndSnapshot(response);
assertEquals(6, savedPages.size());
assertEquals(6, new HashSet<>(savedPages).size());
}
}
@@ -323,4 +329,29 @@ class RearrangePagesPDFControllerTest {
assertEquals(4, realDoc.getNumberOfPages());
}
}
@Test
void testRearrangePages_SideStitchBooklet_RepeatedPaddingPagesAreDistinctNodes()
throws IOException {
MockMultipartFile file = createMockPdf();
RearrangePagesRequest request = new RearrangePagesRequest();
request.setFileInput(file);
request.setPageNumbers("");
request.setCustomMode("SIDE_STITCH_BOOKLET_SORT");
// 6 pages is not a multiple of 4, so booklet padding repeats the last page index
// several times; each repeat must be a distinct page node, not one shared node.
try (PDDocument realDoc = buildRealPdf(6)) {
when(pdfDocumentFactory.load(file)).thenReturn(realDoc);
ResponseEntity<Resource> response = controller.rearrangePages(request);
assertNotNull(response);
assertEquals(200, response.getStatusCode().value());
assertEquals(8, realDoc.getNumberOfPages());
List<Object> savedPages = reloadAndSnapshot(response);
assertEquals(8, savedPages.size());
assertEquals(8, new HashSet<>(savedPages).size());
}
}
}